Kareem Hegazy

Kareem Hegazy

Postdoctoral Researcher

Publications

Neural Equilibria for Long-Term Prediction of Nonlinear Conservation Laws

Published in Under review at ICML, 2026

NeurDE — a ‘neural twin’ of Boltzmann-BGK kinetic theory that pairs exact lattice-Boltzmann transport with a learned equilibrium map — delivers data-efficient, stable long-horizon forecasts of shock propagation and complex compressible dynamics.

Recommended citation: J. Benitez, K. Hegazy, et al. (2026). "Neural Equilibria for Long-Term Prediction of Nonlinear Conservation Laws." Under review at ICML.

Recency Biased Causal Attention for Time-series Forecasting

Published in AISTATS 2026, 2026

Introduces a recency-biased causal attention mechanism that reweights Transformer attention with a smooth heavy-tailed decay — strengthens temporally local dependencies and achieves competitive or superior performance on time-series forecasting benchmarks.

Recommended citation: K. Hegazy, M. W. Mahoney, N. B. Erichson. (2026). "Recency Biased Causal Attention for Time-series Forecasting." AISTATS. https://virtual.aistats.org/virtual/2026/poster/13795

The False Promise of Zero-Shot Super-Resolution in Machine-Learned Operators

Published in ICLR 2026, 2026

Machine-learned operators cannot actually do zero-shot super-resolution — they’re brittle and susceptible to aliasing off the training grid. A simple, data-driven multi-resolution training protocol restores accurate cross-resolution inference.

Recommended citation: M. Sakarvadia, K. Hegazy, et al. (2026). "The False Promise of Zero-Shot Super-Resolution in Machine-Learned Operators." ICLR. https://openreview.net/forum?id=hkF7ZM7fEp&referrer=%5Bthe%20profile%20of%20Kareem%20Hegazy%5D(%2Fprofile%3Fid%3D~Kareem_Hegazy1)

Foundation Models for Discovery and Exploration in Chemical Space

Published in Submitted to Nature, 2025

A foundation-model approach to discovery and exploration in chemical space — unified molecular representations that generalize across diverse downstream tasks in chemistry and materials discovery.

Recommended citation: A. Wadell, A. Bhutani, V. Azumah, et al. (2025). "Foundation Models for Discovery and Exploration in Chemical Space." Submitted to Nature. arXiv:2510.18900.

Neural Network Path Optimization for Finding Transition States on a Machine Learned Potential

Published in Submitted to Nature Computational Science, 2025

A double-ended neural-network framework for chemical transition state search on ML interatomic potentials — outperforms NEB / GSM at high-throughput scale.

Recommended citation: K. Hegazy, E. Yuan, et al. (2025). "Neural Network Path Optimization for Finding Transition States on a Machine Learned Potential." Submitted to Nature Computational Science.

Applying Bayesian Inference and Deterministic Anisotropy to Retrieve the Molecular Structure from Gas-Phase Diffraction Experiments

Published in Communications Physics, 6, 325 (2023), 2023

Novel Bayesian statistical framework that transforms gas-phase diffraction into 3D molecular images — 100× resolution improvement, no simulation bottleneck.

Recommended citation: K. Hegazy, V. Makhija, et al. (2023). "Applying Bayesian Inference and Deterministic Anisotropy to Retrieve the Molecular Structure from Gas-Phase Diffraction Experiments." Communications Physics (Nature), 6, 325. https://www.nature.com/articles/s42005-023-01448-x

Conformer-Specific Photochemistry Imaged in Real Space and Time

Published in Science, 374, 178 (2021), 2021

Directly images conformer-specific electrocyclic ring-opening of α-phellandrene with MeV ultrafast electron diffraction, confirming Woodward–Hoffmann rules in real space and time — a new tool for resolving conformer-dependent reaction dynamics.

Recommended citation: E. G. Champenois, D. M. Sanchez, J. Yang, et al. (2021). "Conformer-Specific Photochemistry Imaged in Real Space and Time." Science, 374(6564), 178-182. https://www.science.org/doi/abs/10.1126/science.abk3132

The Photochemical Ring-Opening of 1,3-Cyclohexadiene Imaged by Ultrafast Electron Diffraction

Published in Nature Chemistry, 11, 504 (2019), 2019

Femtosecond, sub-ångström imaging of 1,3-cyclohexadiene’s photochemical ring-opening with MeV ultrafast electron diffraction — a textbook electrocyclic reaction and model for photobiological reactions such as vitamin D synthesis.

Recommended citation: T. J. A. Wolf, D. M. Sanchez, J. Yang, et al. (2019). "The Photochemical Ring-Opening of 1,3-Cyclohexadiene Imaged by Ultrafast Electron Diffraction." Nature Chemistry, 11(6), 504-509. https://doi.org/10.1038/s41557-019-0252-7

Imaging CF3I Conical Intersection and Photodissociation Dynamics with Ultrafast Electron Diffraction

Published in Science, 361, 64 (2018), 2018

MeV ultrafast electron diffraction images the photodissociation of CF3I through a conical intersection — the first real-space mapping of a coherent nuclear wave packet bifurcating onto two potential energy surfaces as it passes through the intersection.

Recommended citation: J. Yang, X. Zhu, T. J. A. Wolf, et al. (2018). "Imaging CF3I Conical Intersection and Photodissociation Dynamics with Ultrafast Electron Diffraction." Science, 361(6397), 64-67. https://www.science.org/doi/10.1126/science.aat0049